Overview
Market signals detect statistical anomalies in on-chain derivatives data and alert you when conditions that may affect your position are unusual. They are not trade signals — they are context signals.
What signals monitor
Currently live:
| Signal | Data source | What it detects |
|---|---|---|
| Funding Rate | Perpetual futures markets | Funding rate moving outside its 30-day percentile range |
| Open Interest | Perpetual futures markets | OI spike or collapse relative to 30-day baseline |
More signal types are on the roadmap (liquidation cluster detection, exchange net flows, utilization rate spikes).
How the threshold works
Each signal compares the current value against a rolling 30-day distribution of that value. The threshold is expressed as a percentile:
- P95 — signal fires when the metric is in the top 5% of its 30-day range (extreme high)
- P5 — signal fires when the metric is in the bottom 5% of its 30-day range (extreme low)
For example: if the ETH funding rate has averaged between -0.01% and 0.05% over the past 30 days, a funding rate above ~0.045% (top 5%) would trigger a P95 alert.
Alert cadence
Signals are checked every 8 hours. When a signal fires:
- Kairu sends a Telegram notification with the signal type, severity, and your current position context
- The signal appears in the Signals section of the dashboard
- A follow-up check runs at the next 8-hour interval — if the signal has normalised, a resolution notification is sent
Deduplication
If a signal condition persists across multiple checks, Kairu sends a single alert rather than repeating it every 8 hours. A new alert fires only when:
- The condition clears and then re-triggers
- The severity escalates (e.g. P95 → P99)
Signals in AI chat
You can ask the AI about active signals directly:
"Are there any active funding rate signals?"
"What does the current OI signal mean for my position?"
"Is the ETH funding rate historically unusual right now?"
The get_signals tool returns the current signal state with context.
Signals via MCP
get_signals isn't only reachable from Kairu's own chat — it's one of the 13 tools exposed over MCP, so any connected AI client can call it alongside your position and simulation data in the same turn. This is where signals compound with the rest of Kairu: a signal alone tells you the market moved; combined with your position and a simulation, an external AI can tell you what that move means for you specifically, in one query.
For example, in Claude Desktop connected to Kairu:
"Check my active signals. If any are firing, simulate what happens to my
health factor if the underlying asset moves another 15% in the same direction."
That single prompt chains get_signals → get_position → simulate without any Kairu-specific glue — the AI decides which tools to call based on what the signal check returns.